Immich Review

Immich runs on your own hardware and gives you a timeline, albums, and machine-learning search over your photos and videos. Features include face recognition for browsing by person, smart search by natural language and visual similarity, a map view for geotagged photos, shared albums and partner sharing, video support, RAW formats such as CR2, NEF, and ARW, and duplicate detection. Mobile apps are available from the Apple App Store and Google Play, as an APK from GitHub Releases, and through Obtainium and FUTO F-Droid, and you pick which albums to back up, with a Job Queues tab showing processing progress. For migration, the docs link to immich-go for Google Takeout imports, a CLI for uploads, and External Libraries for existing photo archives.

The quick start recommends at least 6 GB of RAM and 2 CPU cores and requires Docker with the Compose plugin (the old docker-compose binary is deprecated and unsupported). The requirements page is more detailed: 6 GB minimum and 8 GB recommended (4 GB works if you disable machine learning), 2 cores minimum and 4 recommended, on amd64 or arm64, and from v3 the amd64 machine-learning container needs the x86-64-v2 microarchitecture level. A Linux system is recommended, and Docker in an LXC container is not. You download docker-compose.yml and example.env, rename the latter to .env, set UPLOAD_LOCATION, DB_PASSWORD, and your time zone, and run docker compose up -d. The web app runs on port 2283 and the first registered user becomes the admin. Postgres data is typically 1 to 3 GB and must be on local SSD storage, never a network share (and on Windows, NTFS, exFAT, and FAT32 do not work, nor do WSL mounts under /mnt). Thumbnails and transcoded video add roughly 10 to 20 percent to your library size.

The web and mobile apps follow the familiar Google Photos layout: timeline, albums, people, places, and search. Machine learning runs on your server, so recognition and search speed depend on your hardware. The docs point out that Immich has built-in database backups but the database does not contain your images and videos. This section reflects the official quick start, requirements, and repository; we did not run an instance.

• Automatic mobile backup for photos and videos, with album selection
• Face recognition and natural-language smart search
• Map view, shared albums, and partner sharing
• RAW formats such as CR2, NEF, and ARW, plus video support
• Duplicate detection and External Libraries for existing archives
• immich-go and a CLI for importing from Google Takeout and bulk uploads
• Built-in database backups and no subscription or storage cap

• The most complete open-source answer to Google Photos, with about 116,000 stars
• Very active development, with v3.3.1 and commits within hours of this review
• Strong migration tooling for Google Takeout and existing libraries
• Full ownership of your photos with no recurring fee

• It is not a backup solution: the database holds only metadata, and you must back up the files in UPLOAD_LOCATION separately
• Needs an always-on server with at least 6 GB of RAM for the full feature set, and Postgres cannot live on a network share
• From v3 the amd64 machine-learning container needs x86-64-v2, and the last release supporting x86-64-v1 is v2.7.5, which is no longer supported
• Docker in LXC and NTFS-backed database storage on Windows are unsupported setups
• AGPL-3.0 applies if you modify and host it for others

Immich is free, with no subscription and no storage cap. Your costs are the server, disks, electricity, and a separate backup of your uploads. The OpenAlternative page also lists a sponsored hosting promotion, which is not Immich pricing.

Google Photos is a polished cloud service with strong search and effortless sharing, but it keeps your library on Google's servers under a subscription once you pass the free tier. Apple Photos is the equivalent inside the Apple ecosystem with iCloud. Immich offers comparable features on your own server, giving you control and unlimited storage in exchange for running the stack and owning backups. The trade-off is that you replace a managed, redundant service with your own hardware and procedures.

• People who want to leave Google Photos or iCloud and own their library
• Families who share albums and want face search without a subscription
• Homelab users already comfortable with Docker
• Anyone who needs to import a large Google Takeout archive

• Anyone who does not want to run, update, and back up a server
• People with low-powered hardware or less than 6 GB of RAM to spare
• Users who depend on Google's cloud-only AI features